- © 2004 American Mineralogist
A total of 30 synthetic samples of the Ca2Fe2−xAlxO5, 0.00 ≤ x ≤ 1.34 solid solution series have been investigated by single crystal X-ray diffraction at 25 °C. Pure Ca2Fe2O5 and samples up to x = 0.56 have space group Pnma, Z = 4, whereas samples with x >0.56 show I2mb symmetry, Z = 4. The substitution of Fe3+ by the smaller Al3+ cation decreases unit-cell parameters and average octahedral and tetrahedral bond lengths and induces distinct changes in the O-atom coordination of the interstitial Ca atom. Discontinuities in the structural parameters vs. the Al3+tot content and changes in slope of these quantities are associated with the phase transition. The essential difference between the two modifications is the cation-O atom-cation angle within the planes of corner sharing octahedra, which is close to 180° in I2mb, but ≈184° in the Pnma phase, and the existence of two different orientations of the tetrahedral chains in Pnma as opposed to one in I2mb. At low overall Al3+ concentrations Al3+ preferentially enters the tetrahedral site until ≈2/3 of it is filled. Additional Al3+ cations, substituted for Fe3+, are equally distributed over octahedral and tetrahedral sites. At high temperature pure Ca2Fe2O5 transforms to a body-centered structure at 724(4)°C. Substituting Al3+ for Fe3+ linearly decreases the transition temperature by 15 °C per 0.1 Al3+ down to 623(5)°C for x = 0.65.
Brownmillerite, Ca2AlFeO5, oxide composition 4CaO·Al2O3·Fe2O3 (C4AF in the nomenclature of cement chemistry), is one of the four most common components of ordinary portland cement clinkers (Taylor 1997). It represents one point in an Al/Fe solid solution series, Ca2(Fe2−xAlx)O5 with x = 1.0. Hansen et al. (1928) were the first to study the phase relations in the CaO-Al2O3-Fe2O3 system. A great deal of phase analytical and structural work has been devoted to the system since then (e.g., Bertaut et al. 1959; Smith 1962) including a detailed review of the crystal chemistry of Al3+-substituted Ca2Fe2O5 by Geller et al. (1971). At atmospheric pressure a continuous solid-solution series exists in the compositional range 0.0 ≤ x ≤ 1.4 (e.g., Smith 1962; Geller et al. 1971; Taylor 1997). The Ca2(Fe2−xAlx)O5 compounds have orthorhombic symmetry, however, they are not isostructural throughout the complete solid solution series. Samples with compositions less than x = 0.6 were found to have space group Pnma (Bertaut et al. 1959; Smith 1962), whereas for samples with x >0.66 some controversy concerning the exact symmetry exists in the older literature. Smith (1962) proposed space group Icmm but subsequently it was discovered that these samples have a slightly modified structure with space group I2mb (Colville and Geller 1971, 1972).
The crystal structure of pure Ca2Fe2O5 was first described by Bertaut et al. (1959) and later re-determined by Colville (1970). In agreement with Smith (1962) it belongs to space group Pnma with a = 5.429(1), b = 14.771(2), and c = 5.599(1) Å. Kahlenberg and Fischer (2000) recently investigated an Al3+-substituted Ca2Fe2O5 sample with x = 0.126, also belonging to space group Pnma. They stated that Al3+ is ordered at the tetrahedral site and no Al3+ is found at the octahedral position. No further structural studies exist for the Pnma phase. Samples with x = 0.57 and 0.72 (Colville and Geller 1972) and brownmillerite Ca2AlFeO5 (Colville and Geller 1971) itself show I2mb symmetry. Pure Ca2Al2O5, which can only be synthesized at high pressure (Aggarwal et al. 1972), also crystallizes in space group I2mb (Kahlenberg et al. 2000). The general structural building units are the same, regardless of space group symmetry, and consists of alternating layers of corner sharing (Fe, Al)O6 octahedra and chains of corner-sharing (Fe, Al)O4 tetrahedra running parallel to a. Ca2+ is located in the interstitial space between octahedral sheets and layers of single-chain tetrahedra. A representation of the structure of pure Ca2Fe2O5 is shown in two different orientations in Figure 1⇓.
Woermann et al. (1968) and Kahlenberg et al. (1997) noticed weak thermal effects in DTA analyses at T1 ≈445 °C and T2 ≈690 °C in pure Ca2Fe2O5 . T1 corresponds to the Néel temperature as Ca2Fe2O5 is magnetically ordered at room temperature (23 °C), whereas T2 represents a phase transition from Pnma to a body-centered cell (Woermann et al. 1968; Kahlenberg et al. 1997). Berastegui et al. (1999) studied the structure of Ca2Fe2O5 as a function of temperature up to 1000 °C by means of powder neutron diffraction and found a phase transition from Pcmn (a different setting of Pnma) to Icmm at about 700 °C. For the high-temperature phase the authors preferred the centric space group Icmm as it gave slightly better profile agreement indices as compared to the acentric space group I2mb. However, their data density is low and no information is obtainable as to whether there are discontinuities in the lattice parameters associated with the magnetic and the crystallographic phase transition. Very recently Fukuda and Ando (2002) determined the Pnma-I2mb phase boundary at high temperature in the Ca2Fe2−xAlxO5 system using X-ray powder diffraction. Pure Ca2Fe2O5 transforms to I2mb at 685 °C; substituting Al3+ into Ca2Fe2O5 decreases the transformation temperature non-linearly (Fukuda and Ando 2002).
Samples of the Ca2(Fe2−xAlx)O5 solid solution series show congruent melting. The melting point of Ca2Fe2O5 is ≈1450 °C and is lowered with increasing Al3+ content to about 1390 °C for samples with x = 1.4 (Newkirk and Thwaitite 1958). Thus single crystals can be prepared by slow cooling of stoichiometric melts (e.g., Smith 1962) or by Czochralski techniques (Colville 1970; Colville and Geller 1971, 1972). Kahlenberg and Fischer (2000) used a flux technique with CaCl2 as a solvent to grow single crystals with x = 0.126 at lower temperatures (1200 °C) in order to avoid possible reduction of Fe3+ to Fe2+, which may take place at temperatures above 1325 °C and atmospheric conditions.
To date all structural investigations of the Ca2(Fe2−xAlx)O5 series have been restricted to two compositions of the Pnma phase (x = 0.0 and 0.126) and three compositions of the body-centered phase (x = 0.57, 0.72, and 1.00). No comparative and systematic crystal chemical study of changes in structural parameters with variation of the Al/Fe ratio has been performed before now. The major objective of this study is to provide this crystal chemical information on a finer compositional grid through the compositionally driven Pnma → I2mb phase transition and to extent it to higher Al3+ concentrations. In addition, no description of changes in bond lengths, bond angles, and distortion parameters, associated with the phase transition, is available to date; this information will be given here. In the present study we also re-evaluate the stability field of the Pnma phase as a function of temperature and Al3+ content and determine the thermal expansion of pure and Al3+-substituted Ca2Fe2O5 as the stability field given by Fukuda and Ando (2002) seems to be incorrect in some respects.
Material synthesis and mineral chemistry
Single crystals were synthesized using both flux growth techniques and slow cooling of melts. Starting materials with the desired stoichiometry were prepared from mechanically homogenized mixtures of CaCO3, Fe2O3, and Al2O3. For flux growth experiments, CaCl2 was added as a solvent to the starting material in a ratio of 1:3. A 4 g batch of the mixture (flux and sample) was placed in a platinum crucible with a platinum cover, slowly heated to 1100 °C in a resistance furnace, and held at this temperature for 24 hours in order to homogenize the melt. The temperature was then lowered from 1100 to 900 °C at a rate of 5 °/hour. After this the platinum crucible was quenched to room temperature by removing it from the furnace and placing it into a ceramic bowl. Crystals can easily be separated from the solidified flux by hand after the synthesis batch has been subjected to atmospheric (humid) air conditions for a few days leading to a hydrolysis of CaCl2. Dissolving the synthesis batch in water is an alternative method to remove the solidified CaCl2 flux, however, this should be avoided since the brownmillerite phases themselves tend to hydrolyze. Flux growth, as described above, was applied to starting materials with x = 0.1, 0.3, 0.6, 0.7, and 1.0 and resulted in prismatic to cuboid, reddish-brown to black crystals with up to 1 mm edge lengths.
Single crystal X-ray diffraction was used to determine the actual amount of Al3+ substituted for Fe3+ at the octahedral and tetrahedral sites by refining the occupation factor of the specific sites. This is straightforward due to the large difference of the atomic scattering factors for Fe3+ and Al3+ and has the advantage (over bulk chemical or microprobe analysis) that it yields site-specific compositional information. A somewhat unexpected result occurred upon analyzing the data for crystals grown by the flux method at comparatively low temperatures and slow cooling: single crystals grown from mixtures with x = 0.3, 0.6, 0.7, and 1.0 in the starting material showed a final Al3+ content of only x ≈ 0.25 throughout the series. It was not possible to obtain samples with different Fe/Al ratios using the applied flux method (samples with prefix BF...). Thus, in a second synthesis series, starting materials with x = 0.65, 0.80, 0.90, and 1.20 were mixed with CaCl2 as a solvent in a ratio of 1:3, placed into covered platinum crucibles and slowly heated to 1200 °C. Instead of slowly cooling the samples, they were isothermally heated at 1200 °C for 7 days and quenched afterward as described above. This procedure again resulted in prismatic to cuboid, reddish-brown to black crystals with edge lengths up to 0.7 mm. Refinement of the actual amount of Al3+ substituted for Fe3+ in these samples from single crystal X-ray intensity data led to values between x = 0.53 and 0.56, larger than before but still without following the nominal Al3+ content. We conclude that it is not possible to obtain a continuous Al/Fe solid solution series for Ca2(Fe2−xAlx)O5 in the presence of CaCl2; a miscibility gap is proposed to be present. More detailed phase analytical investigations are necessary to clarify the limited Al3+ solubility in the presence of CaCl2. This, however, goes beyond the scope of the present study, and will be investigated in a forthcoming paper. It may be noted here that Harchand et al. (1984) report a miscibility gap in the pseudo-binary system CA-CF [Ca(Al2−xFex)O4] in the range 0.25 < x <1.66. The limit for maximum Al-content in C2F (Ca2Fe2O5) and in CF (CaFe2O4) at low temperatures are indeed very similar. However, the results of the present study also suggest that at temperatures at and above 1300 °C this proposed miscibility gap in the Ca2(Fe2−xAlx)O5 series is closed.
Due to the fact that it is not possible to obtain a continuous Al/Fe solid solution by growth from a flux, we decided to synthesize samples directly by slow cooling of the melts. For this, the starting material was placed in a small platinum tube (8 mm in length, inner diameter 3 mm), welded tight on one side with the other side left open. Two such tubes were put into a small corundum crucible, placed into a vertical resistance furnace with a platinum wire, and slowly heated to 1500 °C. After 25 hours of homogenization, the temperature was lowered at a rate of 0.5 °C per hour to 1300 °C. The samples were kept at this temperature for another 22–25 hours and were then rapidly quenched by cutting the platinum wire and dropping the sample into a bowl filled with quartz wool and ice water to allow fast cooling and to freeze the Al/Fe distribution at 1300 °C. By this (rather time consuming) method we obtained a further 21 samples with different Al/Fe ratios (samples with prefix BHT...). The shape of the single crystals ranges from thin-platy to cuboid. When thin-platy, a pronounced pleochroism from deep-red to dark brown is observable under the polarization microscope. The single crystals of pure Ca2Fe2O5 used in this study were obtained by the ceramic sintering method at 1350 °C (sample with prefix BX000).
The chemical composition of selected Ca2(Fe2−xAlx)O5 BHT samples was checked by electron microprobe analysis (JEOL JXA 8600 microprobe, acceleration voltage 15 kV, beam current 30 nA, beam diameter focused to 3 μm). For that purpose crystals were embedded in epoxy resin, polished, covered with carbon and analyzed. At least five analyses from rim to core to rim were measured for each grain and about 20 analyses were performed for each synthesis batch and merged to give the final chemical composition of the sample. The chemical compositions (wt% and structural formula) are reported in Table 1⇓. In all cases the Al3+ contents, determined by microprobe analysis, are in excellent agreement with those obtained from the single-crystal structure refinements.
Prior to the single-crystal growth experiments, several samples of Ca2(Fe2−xAlx)O5 with x up to 1.34 were prepared as polycrystalline powders by solid-state ceramic sintering techniques at 1300 °C and ambient pressure from stoichiometric mixtures of CaCO3, Fe2O3, and Al2O3 in open platinum crucibles. Four cycles of regrinding and re-sintering (for a period of 7 days) were performed in order to guarantee that homogeneous mixed crystals were obtained, even if the desired phases were apparent at the end of the first of the five sintering cycles. The final products are free of additional phases and show sharp Bragg-reflections, which can all be indexed on basis of the brownmillerite metric. The lattice parameters are given in their numeric values as additional material in Deposit Appendix Table 1⇑. 1 The samples produced by ceramic sintering were used for high temperature (HT) powder X-ray diffraction experiments. It may be noted here that the HT study was done some time before any single crystal work and we had no opportunity to repeat the HT measurements with sample material obtained from the single-crystal synthesis experiments described above. As additional material, Deposit Appendix Table 2⇓ lists all the samples used for single crystal X-ray diffraction and Deposit Appendix Table 3 gives the latice parameters for all samples at 25 °C, studied here.
To determine the valence state of iron 57Fe Mössbauer spectra were collected at room temperature with a Mössbauer apparatus in a horizontal arrangement (57Co/Rh single-line thin source, constant acceleration, symmetric triangular velocity shape, multi-channel analyzer with 1024 channels, and velocity calibration to α-Fe). For Mössbauer absorber preparation, approximately one half of the melt-grown material was selected, checked under the optical microscopy for purity, carefully ground under ethanol, and analyzed with X-ray powder diffraction to check for phase purity and lattice parameter refinement. After this the sample material was mixed with powdered sugar and placed in Cu rings (inner diameter 10 mm) covered with a high purity Al-foil on one side. The 57Fe Mössbauer spectra were analyzed with the program suite RECOIL (Rancourt and Ping 1991) in the full static hyperfine interaction Hamiltonian approach using symmetric Lorentzian-shaped subspectra. A complete thickness correction was applied to all data (Rancourt et al. 1993). A detailed temperature-dependent Mössbauer spectroscopic investigation combined with neutron diffraction of the samples is in preparation and will be presented elsewhere.
High temperature X-ray powder diffraction
To obtain detailed information on the temperature variation of the lattice parameters and the high temperature phase transition of the Pnma phase, step-scan X-ray powder diffraction data were collected on the samples synthesized by ceramic sintering in the temperature range 25–800 °C. The high-temperature measurements were recorded on a Θ-Θ Philips X’Pert diffractometer system, equipped with a PAAR HTK-16 high-temperature chamber, and operated with a Cu X-ray tube. To minimize the temperature gradients within the sample in high-temperature experiments, the finely ground sample material was smeared onto the platinum tape as a thin film. X-ray powder diffraction data were recorded between 10 and 120 °2Θ. For temperatures below 600 °C silicon was used as an internal standard. For T >600 °C silicon cannot be used as internal standard because it reacts with the platinum tape; Al2O3 was used instead. Lattice parameters were obtained by whole-pattern refinement using the Le Bail method implemented in the program FULLPROF (Rodriguez-Carajal 2001). To check the real temperature of the platinum tape we performed several initial and intermediate test measurements with silicon and corundum (α-Al2O3) only. For the latter experiments the Bragg reflections of the platinum tape were also clearly observable and thus the unit-cell dimension a of platinum was determinable. From the observed linear variations of the unit-cell parameters we found linear thermal expansion coefficients α = 4.2–4.4 × 10−6/K for pure Si and α = 9.3–9.6 × 10−6/K for platinum in excellent agreement with values given in the literature. For α-Al2O3 the values are α = 8.0–8.1 × 10−6/K for a and α = 8.2–8.3 × 10−6/K for c in reasonable agreement with the values of Aldebert and Traverse (1984) who gave values of 7.3 and 8.3 × 10−6/K for a and c respectively.
Single-crystal structure refinement
Room-temperature single-crystal X-ray diffraction data sets were measured on an imaging-plate diffractometer system (Stoe-IPDS, MoKαradiation, pyrolytic graphite monochromator). Intensity data were collected to 56.5° in 2𝛉 within a φ-range of 0–210°, and the φ increment was 1.5 °/image. The programs X-SHAPE (Stoe and Cie 1996), SHELXS-97 (Sheldrick 1997a), and SHELXL-97 (Sheldrick 1997b) were used for absorption correction, structure solution, and structure refinement respectively. X-ray scattering factors in their ionic form, together with anomalous dispersion coefficients, were taken from the International Tables for Crystallography (Wilson 1992). All lattice parameters were determined from powder X-ray diffraction data (Siemens D500, Cu Kαradiation, 10–120° 2𝛉 range, secondary graphite monochromator, silicon used as internal standard) and were used in the structure refinement. For this, about one half of the single crystal synthesis batch (BX000, BHT..., and BF...samples) was finely ground under ethanol; some of these samples were subsequently used for 57Fe Mössbauer spectroscopy.
Results and discussion
Figures 2a–b⇓ display a typical 57Fe Mössbauer spectrum of a sample of the Ca2Fe2−xAlxO5 series (x = 1.06) in the non-magnetic state. The spectrum consists of two slightly asymmetric resonance absorption lines. Furthermore, the resonance absorption line at +1.20 mm/s shows a slight shoulder (arrow in Fig. 2b⇓) on the low-velocity side. Although two resonance absorption lines are developed, it is not possible to adequately refine the spectrum with only one doublet. Such a refinement strategy yields statistically unsatisfactorily results (χ2 values <4.0) and large misfits are observed, which are most prominent for the +1.20 mm/s absorption line (Fig. 2a⇓). Thus, the two resonance absorption lines correspond to a superposition of two doublets and only a two-doublet model gives acceptable fits. Regardless of the choice of starting parameters, the refinements always converged to the values given in Table 2⇑ within one estimated standard deviation. It is the slight asymmetry of the two absorption lines and the shoulder on the +1.2 mm/s line, which make it possible to find a robust global minimum. The solid curve drawn through the data points is the least-squares fit using two symmetric Lorentzian shaped doublets; the dashed curves represent the two subspectra. Based on the numeric values of the isomer shift δ (Table 2⇑), the low-intensity subspectrum with the smaller δ-value is assigned to high-spin ferric iron at the tetrahedral site and the other to high-spin ferric iron at the octahedral sites. For both sites the quadrupole splitting is among the highest observed for Fe3+. This is indicative of a distinctly asymmetric charge distribution around the probe nucleus resulting from a distorted (elongated) local geometry around the Fe3+ ion at the corresponding octahedral and tetrahedral sites respectively. The quadrupole splitting increases somewhat with increasing Al3+ content. This is an indication of an increasing distortion of the O-atom coordination around ferric iron with increasing Al3+tot. Most important for the purpose of the present paper is the absence of any signs indicative of ferrous iron. This means that during single crystal growth from the melt at high temperatures (1500 °C to 1300 °C) no reduction of ferric to ferrous iron has taken place.
Pure Ca2Fe2O5 and samples up to x ≈1.0 (Geller et al. 1971; Redhammer unpublished results) are (antiferro) magnetically ordered at room temperature. At temperatures below TN, both magnetic dipole and electric quadrupole interactions occur simultaneously. This allows six transitions between the exited and the ground state of the 57Fe probe nucleus and results in a magnetically split 6-line spectrum (instead of a doublet when only electric quadrupole interactions occur). The 57Fe Mössbauer spectrum at 25 °C for pure Ca2Fe2O5 is shown in Figure 2c⇑. The magnetically split spectra of the Ca2Fe2−xAlxO5 series are complex to evaluate, especially when Fe 3+ is partly substituted by Al3+ . Due to the different environments, additional lines appear and line broadening takes place. A subsequent paper will deal with the 57Fe Mössbauer spectra of the Ca2Fe2−xAlxO5 series in detail and we will concentrate on pure Ca2Fe2O5 alone. For this composition, the room-temperature Mössbauer spectrum could be satisfactorily evaluated by making use of two magnetically split 6-line subspectra (Fig. 2c⇑) for high-spin ferric iron at the octahedral and the tetrahedral sites respectively. The point symmetry of the octahedral and tetrahedral Fe3+ sites are 1̅ and m respectively. Thus the electric field gradient (efg) at the octahedral site has no symmetry restrictions, whereas for the tetrahedral site one principal axis has to be perpendicular to the mirror a-c plane, thus restricting the other two efg axes to the a-c plane. The 57Fe hyperfine parameters for Fe3+ in Ca2Fe2O5 below TN are also given in Table 2⇑. The asymmetry parameter η is small for both sites meaning that the efg has axial symmetry for the tetrahedral (η = 0; Vxx= Vyy) and nearly axial symmetry for the octahedral site. The angle 𝛉 between the main component of the efg, Vzz, and the magnetic field at the nucleus, H0, is 90° within experimental error for the tetrahedral and 85° for the octahedral site. This implies that Vzzis oriented in or near a plane perpendicular to H0. As the easy direction of magnetization is parallel to the a axis in Pnma symmetry (Geller et al. 1971), the main efg component Vzzis nearly parallel to the b axis for the octahedral and || to b for the tetrahedral site.
X-ray powder diffraction
Fukuda and Ando (2002) recently determined the temperature-composition (T−x ) stability of the Pnma phase of the Ca2Fe2−xAlxO5 solid solution series and the thermal evolution of the lattice parameters of pure Ca2Fe2O5. Independent of this work, we have determined the boundary of the Pnma phase as a function of T and x and studied the thermal expansion of selected samples along the Ca2Fe2−xAlxO5 series (x = 0.0, 0.40, 0.50, 0.55, 0.70, and 1.00) in more detail. The Bragg reflection (1 3 1), appearing between 29 and 30° 2𝛉 in the Pnma phase with Cu radiation, depending on T and x , can be used to detect the phase transition from Pnma to a body-centered cell at high temperature. Figure 3a⇓ shows a sector of the X-ray powder diffraction pattern displaying the temperature-dependent decay of the intensity of (1 3 1) in Ca2Fe2O5. The X-ray powder diffraction pattern above the phase transition can be refined in space group I2mb or Immb , but the former yields better profile agreement indices.
Along with the disappearance of the h +k +l = 2n + 1 reflections, e.g., (1 3 1), some weak and broad extra reflections appear in the 2𝛉 region between 15° and 30°, namely at 18.9°, 19.8°, 20.8°, 26.1°, and 27.8° (Fig. 3a⇑). These reflections could not be indexed on the basis of orthorhombic symmetry and using the cell parameters obtained from the main reflections of the body-centered phase. However, we can rule out that these reflections are due to an impurity phase or exsolution effects, as the reflections disappear when the temperature is lowered to the region where the Pnma phase is stable (Fig. 3a⇑). The appearance of these extra reflections in the high temperature phase was not mentioned in literature so far. Their presence might be seen as an indication for an incommensurate modulation of the high temperature structure of Ca2Fe2O5 above 725 °C. X-ray diffraction on single crystals is necessary to clarify for this. The symmetry and structure of the high-temperature phase remains an unsettled issue. In what follows, we will assume I2mb symmetry —put in quotations — for the high-temperature phase to indicate that the space group might be wrong. For the following it should be noted further, that this uncertainty in space group assignment only accounts for the powder X-ray diffraction experiments in the high temperature phase, not for the measurements at 25 °C, both in the powder and the single crystal X-ray diffraction experiments. Close to the phase boundary, the intensity of (1 3 1) rapidly drops to zero and completely disappears (Fig. 3b⇑) at the phase transition (Tp). The Pnma -“I2mb” transition takes place at 724 ± 4 °C in Ca2Fe2O5, is completely reversible, and shows no hysteresis behavior within experimental error. All extra reflections observed in the high temperature “I2mb” phase completely disappear The transition temperature in our study is slightly higher than the one found by Fukuda and Ando (2002), Kahlenberg et al. (1997), and Woermann et al. (1968), who proposed a transition temperature of ≈690 °C.
In contrast to the findings of Fukuda and Ando (2002) we found a linear decrease of Tp with increasing Al3+ content (Table 32, Fig. 4⇓) with a sudden drop of the transition temperature above x = 0.65. At room temperature the composition-dependent change from Pnma to I2mb symmetry takes place between x = 0.65 and 0.70. In the latter sample, no indications of a (1 3 1) reflection were present. Also, no extra reflections, which would be similar to those of the high-temperature phase, could be detected. For x = 0.65 a weak (1 3 1) reflection was detectable. For this latter composition, the high temperature X-ray diffraction measurements revealed a transition temperature of 623 ± 5 °C. According to our results, a substitution of 0.1 Al3+ per formula unit lowers the phase transition temperature linearly by ≈15 °C. Woermann et al. (1968) determined the phase boundary of the Pnma -“I2mb” transition for samples of the Ca2Fe2−xAlxO5 series with different x from DTA signals. For x = 0.6 they found a transition temperature of ≈630 °C. A slight increase of the Al3+ content stabilizes the body-centered cell and samples with x >0.6 have I2mb symmetry at 25 °C (Woermann et al. 1968). Smith (1962) found that the phase transition occurs at Al3+ contents of x = 0.66 at 25 °C which is in good agreement with our findings.
Deposit Figures 5a –d3 depict the temperature dependence of the unit-cell parameters for pure Ca2Fe2O5. For a , b and c , as well as for V , there is a linear increase of the unit-cell lengths with temperature up to ≈420 °C. At this temperature there is a discontinuity in the temperature variations. Beyond that the increase of lattice parameters exhibits a slightly greater slope than below 420 °C. Between ≈480 and 700 °C lattice parameters again vary linearly with temperature, and at ≈720 °C a second discontinuity appears, which is most pronounced for the c lattice parameter. The unit-cell volume exhibits both discontinuities in a very distinct way (deposit Fig. 5d). The first kink in the temperature variation of lattice parameters can be correlated with the magnetic phase transition of Ca2Fe2O5, which was found to take place at 440 °C (Geller et al. 1971), whereas the second kink is associated with the Pnma -“I2mb” phase transition. All the other samples, investigated by HT X-ray diffraction, show a linear variation of lattice parameters with temperature. As an example, the temperature variation for Ca2FeAlO5 (x = 1.0) is shown in deposit Figure 63.
We have determined the linear thermal expansion coefficient α, which is given by the formula:
where lT is the unit-cell length at temperature T, l0 is the length at T0 = 25 °C, from the data between 25–420 °C (Pnma -phase beyond the magnetic phase transition) and above 720 °C (“I2mb” phase), The linear thermal expansion of Ca2Fe2O5 was found to be quite anisotropic with the highest value along the b axis (α = 23.1 × 10−6/K), whereas along a and c the linear thermal expansion is nearly one-third of the value along b , namely 8.8 × 10−6/K and 9.3 × 10−6/K, respectively. For the “I2mb” phase (720–800 °C) slightly different values were found (Table 42). It is the b-direction, in which octahedral and tetrahedral layers are stacked up and in which M-Oapex bonds (M = octahedral cation) are pointing (Fig. 1⇑). The large thermal expansion along [0 1 0] may be interpreted in as much as that the tilting of the octahedral layer may decreases towards high temperatures and that M-Oapex bonds expand more rapidly with temperature than the M-O bonds in the equatorial-plane of the octahedra. The thermal expansion along the a -axis, intermediate between the one of b and c , probably is controlled by the expansion of the octahedral site and the changes within the O3-O3-O3 tetrahedral bridging angle, which should change distinctly upon heating. Changes in the tilt of the octahedra will also require changes in the O3-O3-O3 angle. The small thermal expansion along the c -axis is proposed to be controlled by the expansion of the tetrahedral site, which — normally — acts as a very rigid unit and does not show large alterations in bond lengths and bond angles upon heating. The linear thermal expansion shows some variations with composition. Between x = 0.0 and 0.55 it decreases non-linearly with increasing Al3+ content by ≈18% and ≈48% for b and c respectively, whereas the linear expansion coefficient along a increases by ≈20% (Fig. 7⇓, Table 42). The thermal expansion of compositions having I2mb symmetry is different from the Pnma phase. A break in the trend of the data points is observable at the phase transition in Figure 7⇓. Extrapolation of data for samples with I2mb symmetry at 25 °C to pure Ca2Fe2O5 in the “I2mb” symmetry (data between 720 and 800 °C), yields exactly these values measured for Ca2Fe2O5 for lattice parameters b and c .
Single crystal structure refinements of Ca2Fe2−x Alx O5 compounds
From the single crystal X-ray diffraction experiments the change from the primitive to the body-centered cell takes place at x = 0.56 at 25 °C along the Ca2Fe2−xAlxO5 binary join. This is different from the transition composition found in X-ray powder diffraction experiments discussed in the previous section but agrees very well with the findings of Colville and Geller (1971) for samples, grown by the Czrochalski method. They also found a transition composition of x = 0.56. The difference in transition composition may very well be due to the different methods of sample preparation. Analysis of the systematic absences supports the assignment of the centric space group Pnma for all samples with x <0.56. For samples with x >0.56, systematic absences suggested the acentric space group I2mb . Intensity statistics yielded |E2 − 1|= 0.764–0.798, close to the expected value of 0.736 for non-centrosymmetric structures.|E2 −1| steadily increases with increasing Al3+ content. For x <0.56 intensity statistics yield values of |E2 − 1|= 0.923–0.935, and the change of |E2 −1| at the phase transition is very pronounced (Table 5⇓). Experimental details, crystallographic, and refinement parameters of selected samples are given in Table 5⇓, atomic coordinates and equivalent isotropic displacement factors can be found in Table 6⇓, the anisotropic atomic displacement parameters are compiled in deposit Table 72 , and selected interatomic distances and angles as well as selected polyhedral distortion parameters are given in deposit Table 82 . Additional structure refinements with anisotropic atomic displacement parameters in Pnma and I2mb respectively converged to final R1 (all data) values below 4.1% and final w R2 (all data) values below 6.7% (data sets corrected for absorption effects). These maximum values hold true for all compositions studied here, not only to the selection given in Tables 5⇓ through 8. Refinements using the centrosymmetric space group Immb for samples with x >0.56 did not converge to w R2 — values below 35% and no plausible structure models were found.
The equivalent isotropic (Ueq.) as well as the anisotropic displacement parameters (Uij) did not show a systematic variation with the chemical composition of the samples. At best, there may be a slight increase of Ueq with increasing Al3+tot. The M (octahedral) and T (tetrahedral) cations have the smallest Ueq values (0.003–0.006 Å). Calcium and the O1 and O3 atoms have similar Ueq values (0.006–0.009 Å). The largest equivalent isotropic displacements (0.009–0.012 Å) are observed for the O2 atom, which links the octahedral with the tetrahedral sheet.
The lattice parameters of the Ca2Fe2−xAlxO5 solid-solution series decrease with increasing substitution of the smaller Al3+ for the larger Fe3+ ion. The plot of the a unit-cell parameter vs. Al3+ content (Fig. 8a⇓) is a straight line up to x = 0.56, and the same is true for the b unit-cell parameter (Fig. 8b⇓). The c parameter (Fig. 8c⇓) shows only a small, but slightly non-linear decrease with increasing Al3+ . At x ≈0.56 discontinuities appear in all three directions, a , b , and c , and in the unit-cell volume (Fig. 8d⇓). Above this discontinuity point, a again decreases linearly, however with different (greater) slope. The b unit-cell parameter exhibits a non-linear behavior, which is evident at high Al3+ concentrations. The slight jump in lattice parameters upon the transition from the primitive to the body-centered cell and the larger decrease of lattice parameters with increasing Fe3+ by Al3+ substitution is most evident for the c parameter. Literature data, displayed in Figure 8⇓ for comparison, agree with our data for low Al3+ contents. For x >0.6 there is some diversity in the literature and no smooth trends are visible. Our data, however, give rise to smooth variations with compositions, which go beyond what can be learned from the literature. This especially accounts for the slight jump in lattice parameters at the phase transition composition and the course of data at high Al3+tot concentrations. Lattice parameters of the samples synthesized by ceramic sintering at 1300 °C are included in Figure 8⇓ and fit the data trend for the samples synthesized as single crystals. The numeric values of all these lattice parameters can be found in Deposit Appendix Table 1⇑.1
There are two groups of M-O bond lengths within the (Fe3+ , Al3+)O6 site. In Ca2Fe2O5 four out of the six M-O bonds lie in the range 1.961(2)–1.969(2)Å, whereas the other two have lengths of 2.121(2)Å. The latter two equivalent bond distances are those to the O2 atom bridging the octahedral sheet with the tetrahedral chains along b axis. Thus the octahedra appear to be elongated along [0 1 0]. Substituting the smaller Al3+ cation (rVI = 0.530 Å, Shannon and Prewitt 1969) for the larger Fe3+ cation (rVI = 0.645 Å, Shannon and Prewitt 1969) linearly decreases the average M-O distance (Fig. 9a⇓). After the phase transition, there again is a linear decrease of <M-O> vs. Al3+tot, however with a larger negative slope, suggesting that more Al3+ enters the octahedral site. The M-O1 bond lengths (Fig. 9b⇓), interconnecting the octahedra by corner sharing, decrease by 0.5 and 0.7% from x = 0.0 to x = 0.56 (Pnma phase), whereas the M-O2 bond lengths (Fig. 9c⇓) increase slightly by ≈0.3%. At the phase transition a significant jump in M-O1 bond lengths is observed and the course of M-O2 vs. Al3+tot data changes the slope direction. For x >0.56, all three non-equivalent M-O bond lengths decrease with increasing Al3+tot content, the M-O1 bonds by 1.5 and 1.7%, the M-O2 bond by 0.7%. Because of the non-uniform decrease of M-O bond lengths, the bond length distortion (BLD, Renner and Lehmann 1986) of the octahedral site increases with increasing Al3+tot content (Fig. 9d⇓).
The Pnma-I2mb phase transition is also reflected in the aver- age of the O-O distances, defining the edges of the (Fe3+, Al3+)O6 octahedron (deposit Fig. 10a3). As the point symmetry of the octahedron changes from 1̅ (Pnma) to 2 (I2mb), the number of non-equivalent O1-O1 distances within the equatorial plane of the octahedron changes from 2 in Pnma to 3 in I2mb . The M cation, located at (0, 0, 0) in Pnma , is slightly displaced along a in I2mb (x, 0, 0) thus allowing the two equivalent O1-O1 distances [2.840(2)Å] in Ca2Fe2O5 to become inequivalent. Together with the displacements of the O1 atom, this leads to a rather large change in lengths for these O1-O1 octahedral edges at the phase transition (deposit Fig. 10b3). The O1-O2 interatomic distances defining the edges between equatorial O atoms and the apex O atoms of the octahedra display only small composition-dependent changes (decrease and increase) for the Pnma phase (Fig. 10c), consistent with the small increase observed for the M-O2 bond length (deposit Fig. 10a3). Linear decreases with increasing Al3+tot content, however with different slopes, take place for the O1-O2 edge lengths in the body-centered phase (deposit Fig. 10c3). The edge length distortion (ELD) parameter (deposit Fig. 10d3) as defined by Renner and Lehmann (1986) increases from Ca2Fe2O5 to samples with x ≈ 0.56. The phase transition is associated with a sudden increase of ELD, which then slightly decreases with increasing Al3+tot content for the body-centered phase.
Octahedral bond angles O-M-O range between 86.6(1)° and 93.4(1)° in Ca2Fe2O5. The incorporation of Al3+ slightly modifies the O-M-O bond angles. The changes are moderate and less than ≈1° over the complete solid solution series. At the phase transition, discontinuities are present for all non-equivalent bond angles; the situation is shown for the O1-M-O2 bond angle as an example (deposit Fig. 11a3). The quadratic octahedral angle variance OAV, as defined by Robinson et al. (1971), exhibits this discontinuity in a pronounced way (deposit Fig. 11b3). In terms of the OAV, samples with x ≈0.6 have the greatest deviation from ideal octahedral geometry. Increasing the Al3+tot content decreases the bond-angle distortion. As depicted in deposit Figure 11b3 the O1-M-O2 bond angles approach the ideal value of 90 ° with increasing Al3+ substitution. This is also observable for the other O-M-O bond angles (cf. deposit Table 82) and reflects the substitution of the 3d cation Fe3+ by the closed shell cation Al3+ , having different covalence character (directed to non-directed bonds).
As outlined in Figure 1⇑, octahedra are tilted against each other by rotation about [1 0 0] and [0 1 0]. In Ca2Fe2O5 the octahedral are rotated by 9.8° around [1 0 0] and by 3.1° around [1 0]. With increasing Al3+tot content, the tilt around [0 1 0] (Fig. 12a⇓) decreases by ≈1° within the Pnma phase. Similar to perovskite, the tilt is due to the fact that the octahedra are too large for the cation (here Ca2+) within the interstitial site. Reducing the overall size of the octahedra (by substituting the smaller Al3+ cation for Fe3+) also reduces the octahedral tilt. The phase transition is associated with a sudden decrease of the octahedral tilting by ≈0.5 °. In samples with I2mb symmetry, the tilting around b continues to decrease with increasing Al3+tot down to ≈7.5 °. For samples with Pnma symmetry the tilting around [1 0 0] also decreases with increasing Al3+tot content. The discontinuity at the phase transition is even more pronounced (Fig. 12b⇓). In contrast to the tilt around [0 1 0], no significant reduction of tilt around [1 0 0] is observed for the I2mb phase.
In Ca2Fe2O5 the tetrahedral bonds show lengths of 1.843(2) Å for the T-O2 and 1.912(2)-1.914(2) Å for the T-O3 bonds. The shorter T-O2 bonds are those connecting the tetrahedral chain with the octahedral layer, while the T-O3 bonds are those within the tetrahedral chain running along [1 0 0]. With increasing Al3+ substitution, T-O bonds decrease distinctly following a linear trend (deposit Figs. 13a and 13b3) for the Pnma phase. At the phase transition a small jump toward lower values can be seen for the T-O3 bonds. For the I2mb phase, the variation of T-O bonds exhibit a different slope compared to the samples of the Pnma phase. A non-linear course of the T-O2 vs. Al3+tot (deposit Fig. 13b3) and of the <T-O> vs. Al3+tot data is observable at high Al3+ contents (deposit Fig. 13c3).
The O-O interatomic distances in Ca2Fe2O5, defining the edges of the tetrahedron, can also be divided into two groups. Three out of the four are within 3.008(1) and 3.048(2) Å (Fig. 14a⇓), but the fourth distance (O2-O2) is distinctly longer at 3.229(2) Å (Fig. 14b⇓). This is the tetrahedral edge, connecting the two octahedral layers along [0 1 0]. As for the T-O bonds, tetrahedral edges vary linearly with increasing Al3+tot content in the Pnma phase and exhibit some compositionally dependent non-linear behavior at high Al3+ concentrations.
The average O-T-O angle is close to the ideal value of 109.4 ° with no significant compositionally dependent variation. The individual bond angles, however, range between 105.4(1)° and 122.2(1)° in Ca2Fe2O5. The largest value occurs for the angle between the tetrahedral cation T and the two O2 atoms shared by the tetrahedral chain and two neighboring octahedral sheets. This corresponds to an elongation of the tetrahedron along b. The large O2-T-O2 angle is consistent with the large O2-O2 edge length (Fig. 14a⇑). Substituting Fe3+ by Al3+ causes a non uniform variation of the O2-T-O2 angle. For the Pnma phase (Fig. 15a⇓) it decreases linearly with increasing Al3+tot content. In the I2mb phase this trend continues up to x ≈0.75. Beyond this Al3+tot content the slope changes sign and the O2-T-O2 angle increases with increasing Al3+tot content. This behavior is related to the reduced decrease of the O2-O2 tetrahedral edge. As the T-O2 bond continuously decreases, but the change of the O2-O2 bond becomes non-linearly reduced at high Al3+tot contents, the course of the O2-T-O2 bond angle vs. Al3+tot data have to show a positive slope at high Al3+tot contents. The O3-T-O3 angle (Fig. 15b⇓) increases, first slightly non-linearly in Pnma , and, after a small discontinuity at the phase transition, linearly in I2mb . This increase in O3-T-O3 indicates that increasing Al3+ stretches the distorted tetrahedron slightly in the chain direction, which is parallel to the a axis. One of the two O2-T-O3 angles (Fig. 15b⇓) decreases with increasing Al3+, the others increase at the same extent within the Pnma phase. For the I2mb phase both angles decrease. The large spread in tetrahedral bond angles is also observed in large values of the quadratic tetrahedral angle variance TAV (Fig. 15c⇓) as defined by Robinson et al. (1971). For pure Ca2Fe2O5 this amounts to 39.6 °. The approach of O-T-O angles to more ideal values (decrease of O2-T-O2, increase of O3-T-O3 and one O2-T-O3 angle) lowers the TAV within the Pnma phase. It is mainly the inflection of O2-T-O2 vs. Al3+tot data which causes the TAV to increase again at high Al3+tot contents within the I2mb phase. The tetrahedral chains are strongly kinked, expressed by low O3-O3-O3 angles (Fig. 15d⇓). Increasing Al3+ causes an increase of this angle, i.e., the tetrahedral chains become more stretched. This stretching is in close relationship with the decreasing tilting of the octahedra.
Interstitial site (Ca2+).
The Fe3+ by Al3+ substitution primarily affects the octahedral and the tetrahedral site, however it is also reflected in the geometry of the Ca site. The calcium ion is surrounded by a irregular polyhedron of eight O atoms at distances ranging from 2.323(2) to 3.000(2) Å. Figure 16⇓ gives a comparison of the next neighbor geometry of the Ca2+ cation in Pnma (x = 0.00) and I2mb (x = 1.34) symmetry respectively. The mean of the eight different Ca-O bond lengths decreases linearly with Al3+ for both the Pnma and the I2mb phase, however with different slopes and a slight jump at the phase transition (Fig. 17a⇓). The individual Ca-O bond lengths show different compositional variation in Pnma and I2mb . The change in octahedral tilting distinctly alters the O1 environment around the Ca atom and thus is closely related to the changes in individual Ca-O1 bond lengths. The rapid reduction of the octahedral tilt at the phase transition causes rapid changes in Ca-O1 bond lengths (Fig. 17b–c⇓), most evident for the long Ca-O1 bond at 2.732(2) Å in Ca2Fe2O5 (Fig. 17c⇓). A smaller and nearly continuous reduction by 0.7% for the complete compositional range is observed for the Ca-O2 bond at 2.323(2) Å in Ca2Fe2O5. The Ca-O2 bonds connect the Ca2+ cation to both the tetrahedral chain and the octahedral sheet. The Ca-O2 bond at 2.542(2) Å in Ca2Fe2O5 increases by 0.9% within the Pnma phase, while the Ca-O2 bond at 3.000(2) Å in Ca2Fe2O5 decreases by 2.4% within the same compositional range. For the I2mb phase all Ca-O2 bonds decrease with increasing Al3+ . These changes are related to changes in the buckling (kinking) of the tetrahedral chain. The Ca-O3 bond (Fig. 17d⇓), connecting the Ca2+ cation to the bridging O3 atom of the tetrahedral chain, remains rather constant over the whole compositional range. The jump in octahedral tilting and within the Ca-O1 bond lengths at the phase transition is related to the discontinuous movement of the Ca(x) and O1(x) and O1(z) fractional atomic co-ordinates. The above mentioned co-ordinates as well as Ca(z) change steadily increases with increasing Al3+ up to the phase transition, at higher Al3+ contents their numeric values stay nearly constant, correlating with an almost constant tilt angle around [1 0 0]. The reduction of the tilt with increasing Al-content is directly related to a stretching of the tetrahedral chains along the a-axis. Tetrahedral chain and octahedral layer are connected to each other by the O2 oxygen atom. The kinking of tetrahedral chains decreases by ≈5.5 ° from Ca2Fe2O5 to Ca2Fe0.64Al1.34O5, but the chains remain buckled also in I2mb . Similar to perovskite, the underlying reason for the tilt of the octahedra can be found in the size of the Ca2+ cation in the interstitial site of the structure. The movement of the O1 oxygen atom in x, y and z directions gives rise to a symmetric nearly rectangular arrangement of octahedra within the a –c plane in I2mb , while, in Pnma (Fig. 16⇓), the arrangement is much more distorted. Projected onto the a-c plane, the O1-O1-O1 angle is 179.0(1)° along the a axis and exactly 180° along the c – axis in I2mb . One further distinct difference between the Pnma and the I2mb structure concerns the tetrahedral chain. Whereas the chain of tetrahedra at x/a = ¼ are essentially the same in both structures (Fig. 18⇓), the chains at x/a = ¾ are related to each other by an inversion center (−1) in Pnma and by a twofold axis is I2mb . Thus the tetrahedra have a different orientation along the b -axis and build up a stacking sequence of O-T-O-T .-... for Pnma , whereas it is O-T-O-T-... for I2mb .
By refining the site occupation number for the octahedral and tetrahedral site it was possible to determine the cation distribution of Fe3+ and Al3+ over the corresponding sites. Our results show that the distribution is far from random and that Al3+ strongly prefers the tetrahedral site (Fig. 19⇓). How-ever, even at low Al3+ contents, Al3+ enters the octahedral and the tetrahedral site for both types of samples, those grown directly from the melt and quenched from 1300 °C (BHT samples) as well as for samples grown with flux methods and quenched from 900 °C. Mössbauer spectroscopy has been used to check the cation distribution found in the present study by site occupation refinement from single crystal intensity data. The results of the two different methods are in good agreement: sample BHT140 (x = 1.36) was found to contain 0.103(5) and 0.144(7) Fe3+ atoms per formula unit at the tetrahedral site, as determined by site-occupation refinement of single crystal X-ray intensity data and by Mössbauer spectroscopy, respectively. Similar agreement was found for sample BHT130 (x = 1.26) with 0.147(4) and 0.166(8) Fe3+ apfu and sample BHT120 (x = 1.078) with 0.238(5) and 0.235(11) Fe3+ apfu. The cation distribution obtained for our samples is in good agreement with that of Colville and Geller (1971, 1972). Their data are indicated by open crossed squares in Figure 19⇓. Smith (1962) proposed that Al3+ preferentially enters the tetrahedral site until half the tetrahedra are filled, after which additional Al3+ and Fe3+ atoms distribute equally. Figure 19⇓ shows that this view is supported by our data. Above x ≈0.9, corresponding to ≈0.65 Al3+ apfu at the tetrahedral site, the trend of the data points for Al3+tot vs. Al3+oct follows a straight line with a slope that is almost identical to that for an equal distribution of Al3+ and Fe3+ over octahedral and tetrahedral sites respectively. These findings show that above an ≈⅔ filling of the tetrahedral site by Al3+ additional Al3+ , which substitutes for Fe3+ , distributes equally over the octahedral and the tetrahedral site. The non-linear variation of especially T-O2 and <T-O> at high Al3+ contents can be ascribed to the change in Al3+ site preference, i.e., the equal distribution of additional Al3+ over M and T sites. A reason for the change of symmetry from Pnma to I2mb may also be found in the preference of Al3+ for the tetrahedral site at low overall Al3+ contents. As the tetrahedral chains incorporate more Al3+ than the octahedral layer, the tetrahedra shrink and chains can straighten to produce the higher I2mb symmetry. A second source of information on the Fe/Al-distribution can be found in the average bond lengths. The relative change of average octahedral and tetrahedral bond lengths within the Fe/Al binary join is markedly different at low overall Al3+ contents (Fig. 20⇓). Above x ≈0.9 the two graphs (changes of average T-O and aver-age M-O bond lengths relative to Ca2Fe2O5) are almost parallel, supporting the above proposal.
G.J.R. wishes to thank the Austrian Academy of Science for financial support throughout an APART (Austrian Program for Advanced Research and Technology) scholarship during 2000–2003. The technical support of J. Ernst (Aachen) for the synthesis and of V. Kaiser (Aachen) in the X-ray laboratory is gratefully acknowledged. The careful and thorough review and the helpful comments of G. Cruciani and an anonymous referee as well as those of S. Quartieri are gratefully acknowledged.
↵1 Deposit Appendix Tables 1 through 3 are available as deposit item AM-04-059 on the American Mineralogist web site at http://www.minsocam.org. Structural data are available both as crystallographic information files (CIFs) and as a file suitable for the program XtalDraw.
↵2 For a copy of Tables 3, 4, 7, and 8, document item AM-04-057, contact the Business Office of the Mineralogical Society of America (see inside front over of recent issue) for price information. Deposit items may also be available on the American Mineralogist web site at http://www.minsocam.org.
↵3 For a copy of Figures 5, 6, 10, 11, and 13, document item AM-04-058, contact the Business Office of the Mineralogical Society of America (see inside front over of recent issue) for price information. Deposit items may also be available on the American Mineralogist web site at http://www.minsocam.org.
3 Corresponding address: Neuhofen im Innkreis 224/1, A-4910 Ried im Innkreis, Austria
Manuscript handled by Simona Quartieri
- Manuscript Received May 27, 2003.
- Manuscript Accepted September 17, 2003.